If the content type is null, the call for .toString on it will cause an NPE. This fixes this issue by wrapping the call in an Option, then mapping toString or falling back to null. getCharSetFromContentType handles null by returning ISO-8859-1
I noticed this was showing up in profiles when SBT's task engine
was using Keys, etc (that contain Configurations) in HashMap's.
Let's cache it instead. I don't think there is a need to use a lazy
val for this, we can compute it eagerly.
* changes
* changes to Attributes
* make changes
* add test and fix bug
* add more tests
* fix failing tests
* fix parentheses
* remove comments I added and add test that's guarenteed not to exist in a repo
* remove Attributes third parameter and refactor Parse's moduleVersionConfig to return a tuple of Dependency and a Map of String to String representing extra parameters for the dependency
* update some return types and methods in the Helper class
* return FallbackDependenciesRepository back to original state
* refactor helper
* remove url from attributes
* fix tests and add cache to extra in build.sbt
* remove FallbackDepsRepo from sbt-coursier
* add variable for url and call encode in tests
* update tests and helper to have proper behavior
* remove setting deps to intransitive when url is present and add two tests
* add more tests, implement @wisechengyi's comments
* fix nits
* update ParseTests because some failed
* fix tests
* incorporate feedback from @alexarchambault and @wisechengyi
* update ParseTests to check for returned error vs thrown error
* remove one test that is covered in ParseTests
* fix nits
* add back deleted brackets
* return errors in Left without using return statement
* revert change because it's broken
* fix some positional things
* add return statement or else error won't be processed
* fix nits. thanks @wisechengyi
* Remove return statements
Make sure that we generate valid ID (according to matches(ID, s)) so
that we properly test the new validID implementation.
And that's what led to the bug fix. :)
This showed up in profiling. It's known that TypeTags are expensive. Even
more so if the reflect universe is accessed during startup when the
class loading and JIT compiler are busy enough with other stuff.
It turned up in profiling sessions. Previously, it used parser combinators
which are somewhat slow especially when the JVM is still cold.
The grammar for ID is simple enough to afford this handwritten parser.
I've reimplemented java version detection as a bash function.
This no longer uses grep.
Also this no longer uses `?` in sed, which doesn't work on macOS.
Fixes https://github.com/sbt/sbt/issues/3873
Yi Cheng
Ståle Undheim
Dale Wijnand
Alexandre Archambault
Jason Zaugg
Alexandre Archambault
Dorothy Ordogh
xuwei-k
exoego
eugene yokota
Dale Wijnand
Deokhwan Kim
Johannes Rudolph
Jonas Fonseca
Friedrich von Never